Charlie Ashton, Senior Director Business Development at Napatech, looks at the importance of power and cost in edge data centers – the process of offloading virtual switching from server CPUs onto programmable Smart Network Interface Cards (SmartNICs).
For those of you have followed my articles, webinars and video interviews over the last half-year or so, you would have seen me comment on the fact that NFV has been in the trough of disillusionment or the doldrums, as I have described it. I have taken no satisfaction in the fact that many have agreed with me on this point. So, it was with some hope and anticipation that I attended the NFV World Congress in San Jose recently. I wanted to see if we were emerging from trough and starting to make progress.
The short answer is maybe. I see signs of progress, but I note that there are still many that disagree and see many challenges. The question remaining is whether we have turned a corner and that these challenges can be met and overcome or if these challenges are still fundamental and could still prove to be the death-knell of NFV.
Here is why I am optimistic.
It has been pretty clear from the beginning of NFV that achieving its original promise would require then implementing a Management and Orchestration (MANO) framework that would enable service agility and continuous optimization. It would also require a common approach from one carrier to another to allow interoperability.
From this perspective, MANO is a fundamental make-or-break issue. If it is not solved, then NFV is dead in the water, or in the doldrums, if you like, where NFV has been for some time. It is no wonder that this is the case, as it is a totally new paradigm for how carriers operate and get jobs done. It is more than just technology, it is at the very heart of the transformation that carriers need to make. It is about more than technology, it is about culture and organizational change.
My optimism springs from the fact that consensus is forming on how MANO frameworks should look and the kind of interfaces that are needed. The MEF Lifecycle Service Orchestration initiative is delivering on this need with wide backing. ONAP and OpenMANO are providing open implementation alternatives. So, the pieces are coming together to make MANO possible.
But, challenges remain; the non-technical cultural and organizational challenges of implementing MANO, and NFV in general, are now the main stumbling blocks. Formal presentations and statements by carriers at the show highlighted this point while conversations with vendors to those carriers confirmed that these are real challenges.
But, the question is whether these challenges are show-stoppers. I don’t think so. Many of those same presentations highlighted the progress that was being made as well as the target use cases being implemented. Carriers showed a distinct determination to move forward and a willingness to find a way, no matter what. This includes embracing open-source, even to the extent of making open-source the first choice, in order to maintain control and flexibility as well as to accelerate development cycles by avoiding internal procurement and approval processes. This is a good example of carriers getting out of their own way to make NFV a success.
Conversations with vendors also revealed that real deployments are starting to happen and that they are seeing the first real orders.
So, while there are multiple challenges remaining, I fundamentally believe that a significant corner has been turned and that we are seeing the first winds fill the sails of NFV that will propel us out of the doldrums.
What to expect next: focus on insight and automation
So, now that MANO is being set on the right track, the collective minds of the NFV community are starting to think of the next challenge on the horizon. As I stated above, the motivation for MANO is to enable service agility and continuous optimization. However, this is hard to achieve without a NFV infrastructure that can support automation and provide insight (see my previous blog for more on this).
This was also evident at the show with a greater focus on performance and fault management, security and the need for data on how the network is performing so MANO can proactively take action. There is a growing awareness that a rethink of NFV infrastructure is required in order to provide the insight into what is happening in the network.
For this reason, many are now actively engaged in assessing what FPGA SmartNICs can offer in addressing these needs. The term “SmartNIC” encompasses many different types of solutions based on a variety of technologies and I will leave it to a later blog (and possibly a more technically accomplished writer) to tackle that subject. This is both good and bad. Good, in the sense that there is a recognition that alternative approaches to solving the NFV infrastructure performance and efficiency issues are valuable. Bad, in the sense that some of these approaches, which could lead to more overall cost and inefficiencies, could end up with all SmartNIC approaches being tarnished with the same brush.
As an example, one carrier at the show commented on the fact that they were not interested in assessing FPGA SmartNICs any longer as their experiences so far had shown that while the FPGA SmartNICs they assessed provided better performance, they did so in a way that lead to more operational complexity. The performance gains and associated cost improvements they generated were simply dwarfed by the costs of operating these solutions.
What is unfortunate in this case is that this carrier has now made a blanket decision not to consider alternative SmartNIC implementations because of a bad experience with one of these approaches.
My message is simple in this regard; don’t treat all FPGA SmartNICs as equal, as there can be significant differences in how solutions are implemented and their impact on operations. Each SmartNIC should be evaluated on its own terms. At the same time, SmartNIC vendors need to ensure that they are addressing challenges in a way that does not undermine the overall NFV solution and business case (see my earlier blog on solving micro-problems but creating macro-problems).
Nevertheless, it is heartening to see that FPGA SmartNICs which were categorically ruled out of NFV just a few years ago are now being taken seriously as potential solutions to the challenges that the NFV infrastructure has to address.
We, at Napatech, are focused on ensuring that our virtualization solutions based on our FPGA SmartNICs not only efficiently address performance, cost and flexibility needs, but do so in a way that does not compromise the overall business case by hindering continuous optimization, automation or introducing operational complexity. At the same time, we are ensuring that our solutions will support the needs for insight that can drive service agility and continuous optimization so the benefits of recent MANO breakthroughs can be fully realized.